[sldev] Re: [VIEWER] Reducing the LoD is not the same as billboarding.

[Argent Stonecutter]

From: John Hurliman <jhurliman at wsu.edu>
> Argent Stonecutter wrote:
>> From: Mike Monkowski <monkowsk at watson.ibm.com>
>>> For the scene polygon count, I assume you mean the added polygons  
>>> for
>>> the avatars.  This could be handled with lower LOD for more distant
>>> avatars.

>> This has already been fiddled with and the results were horrible.

> That's a pretty quick dismissal of a very broad optimization  
> technique,
> especially since it is going to be introduced again shortly.

Reducing the LoD for avatars at a lower distance than is already done  
is not "avatar billboarding".

Billboarding in general is a tremendously useful technique, and  
really needs to be performed on a wider scale than simply avatars  
(for example I've suggested that the already existing world map  
images could be used to replace the view of the regions at high  
altitudes, letting them act as billboards for the sims and giving a  
useful view of the world for flying vehicles). It's a completely  
different approach from reducing the level of detail of the mesh.

To illustrate, consider https://jira.secondlife.com/browse/VWR-1623.  
In the first three images in that example the level of detail of the  
mesh is being reduced, from a multi-sided prism to a 4-sided prism to  
a 2-sided prism (or plane), without changing the number of elements.  
The fourth image shows a billboarded tree. The second reduction in  
mesh detail produces a result MUCH worse than a billboard, and yet  
costs more to render.

I'd like to see snapshots of whole sims be used as billboards. You  
could do this progressively and get the equivalent of kilometers  
worth of draw distance at a lower cost than 256 meters now.